Forecasting the term structure of natural gas constant maturity futures using a seasonal principal components-based post-weighted monte carlo simulation

ABSTRACT

Given the increasing demand for natural gas, a number of industry participants have attempted to value their storage contracts. Such valuation involves both a forecasting of forward prices, as well as an optimization of spreads and/or options. Leading industry firms have utilized Principal Component Analysis and simulation to develop future values for forward contracts, using a covariance matrix of forward returns generated by historical data. We developed an improved procedure for applying the evolution of the historical covariance matrix. Our methodology involves the determination of a set of weights and the ensuing application of those weights on the results of Monte Carlo simulations under a variable-period switching of the VCV matrices. Out-of-sample results significantly improved the approximation of actual values of forwards and spreads. Furthermore, its deviations from market prices varied considerably less than those resulting from current techniques, allowing future users to calibrate their models to the market easier.